INDEX

Explanations

instances of words related to reality, truth, and tangible attributes

oai_token-act-pair · gpt-3.5-turbo

references to veracity and meaningfulness in contexts

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cius

-0.77

sometimes

-0.72

 Beware

-0.71

alian

-0.69

zed

-0.68

dm

-0.68

ARC

-0.67

 Originally

-0.66

ften

-0.66

almost

-0.64

POSITIVE LOGITS

 whatsoever

1.18

 explanation

1.13

 reason

1.05

 indication

1.01

 distinction

0.96

 recourse

0.90

 reperc

0.90

 affiliation

0.88

nor

0.87

 motive

0.87

Activations Density 0.182%

instances of words related to reality, truth, and tangible attributes

references to veracity and meaningfulness in contexts

No Comments

No Known Activations